Structure design of transparent ceramic armor based on machine learning and particle swarm optimization method

Transparent ceramic armor with laminated structures encounters significant design challenges in achieving an optimal balance among ballistic protection, weight, optical clarity, and cost through thickness optimization. Traditional experimental and simulation-based approaches face difficulties in multiobjective optimization due to their high computational demands and inability to reconcile conflicting requirements. This study introduces a novel machine learning-guided particle swarm optimization framework, representing an advancement in armor design methodology. For a “sapphire/glass/polycarbonate (PC)” armor system subjected to 12.7 mm armor-piercing incendiary (API) threats, we first develop a physics-based Defense function that integrates penetration resistance (residual projectile energy) and protection redundancy (bulge deformation) into a single quantifiable metric on a 0–1 scale. A validated finite element model generates ballistic performance data for 196 configurations, enabling comparative training of three machine learning models. The support vector regression (SVR) model achieves exceptional accuracy (R² = 0.98, rRMSE < 4%) in predicting Defense values, surpassing both XGBoost and neural networks. By integrating this predictive capability with an enhanced multiobjective particle swarm algorithm, we established a real-time feedback optimization framework. This framework simultaneously reduces thickness by 22.2% and enhances transparency by 42.3%, while only incurring a 28.8% increase in cost and ensuring the protection threshold remains above the required level (Defense ≥ 0.5) during the optimization process. Experimental validation demonstrates that the optimized configurations preserve the predetermined ballistic resistance performance while achieving balanced improvements in thickness reduction, transmittance increase, and cost efficiency. This study exemplifies the synergistic potential of machine learning and particle swarm optimization for transparent armor design, offering a generalized approach to multiphysics optimization in protective material systems. It addresses longstanding challenges in traditional trial-and-error methodologies through intelligent algorithms that incorporate constraint-aware thickness distribution.